my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	---
#2	title: Google ADK
#3	---
#4
#5	Integrate [Mem0](https://github.com/mem0ai/mem0) with [Google ADK (Agent Development Kit)](https://github.com/google/adk-python), an open-source framework for building multi-agent workflows. This integration enables agents to access persistent memory across conversations, enhancing context retention and personalization.
#6
#7	## Overview
#8
#9	1. Store and retrieve memories from Mem0 within Google ADK agents
#10	2. Multi-agent workflows with shared memory across hierarchies
#11	3. Retrieve relevant memories from past conversations
#12	4. Personalized responses based on user history
#13
#14	## Prerequisites
#15
#16	Before setting up Mem0 with Google ADK, ensure you have:
#17
#18	1. Installed the required packages:
#19	```bash
#20	pip install google-adk mem0ai python-dotenv
#21	```
#22
#23	2. Valid API keys:
#24	- [Mem0 API Key](https://app.mem0.ai/dashboard/api-keys)
#25	- Google AI Studio API Key
#26
#27	## Basic Integration Example
#28
#29	The following example demonstrates how to create a Google ADK agent with Mem0 memory integration:
#30
#31	```python
#32	import os
#33	import asyncio
#34	from google.adk.agents import Agent
#35	from google.adk.runners import Runner
#36	from google.adk.sessions import InMemorySessionService
#37	from google.genai import types
#38	from mem0 import MemoryClient
#39	from dotenv import load_dotenv
#40
#41	load_dotenv()
#42
#43	# Set up environment variables
#44	# os.environ["GOOGLE_API_KEY"] = "your-google-api-key"
#45	# os.environ["MEM0_API_KEY"] = "your-mem0-api-key"
#46
#47	# Initialize Mem0 client
#48	mem0 = MemoryClient()
#49
#50	# Define memory function tools
#51	def search_memory(query: str, user_id: str) -> dict:
#52	"""Search through past conversations and memories"""
#53	# For Platform API, user_id goes in filters
#54	filters = {"user_id": user_id}
#55	memories = mem0.search(query, filters=filters)
#56	if memories.get('results', []):
#57	memory_list = memories['results']
#58	memory_context = "\n".join([f"- {mem['memory']}" for mem in memory_list])
#59	return {"status": "success", "memories": memory_context}
#60	return {"status": "no_memories", "message": "No relevant memories found"}
#61
#62	def save_memory(content: str, user_id: str) -> dict:
#63	"""Save important information to memory"""
#64	try:
#65	result = mem0.add([{"role": "user", "content": content}], user_id=user_id)
#66	return {"status": "success", "message": "Information saved to memory", "result": result}
#67	except Exception as e:
#68	return {"status": "error", "message": f"Failed to save memory: {str(e)}"}
#69
#70	# Create agent with memory capabilities
#71	personal_assistant = Agent(
#72	name="personal_assistant",
#73	model="gemini-2.0-flash",
#74	instruction="""You are a helpful personal assistant with memory capabilities.
#75	Use the search_memory function to recall past conversations and user preferences.
#76	Use the save_memory function to store important information about the user.
#77	Always personalize your responses based on available memory.""",
#78	description="A personal assistant that remembers user preferences and past interactions",
#79	tools=[search_memory, save_memory]
#80	)
#81
#82	async def chat_with_agent(user_input: str, user_id: str) -> str:
#83	"""
#84	Handle user input with automatic memory integration.
#85
#86	Args:
#87	user_input: The user's message
#88	user_id: Unique identifier for the user
#89
#90	Returns:
#91	The agent's response
#92	"""
#93	# Set up session and runner
#94	session_service = InMemorySessionService()
#95	session = await session_service.create_session(
#96	app_name="memory_assistant",
#97	user_id=user_id,
#98	session_id=f"session_{user_id}"
#99	)
#100	runner = Runner(agent=personal_assistant, app_name="memory_assistant", session_service=session_service)
#101
#102	# Create content and run agent
#103	content = types.Content(role='user', parts=[types.Part(text=user_input)])
#104	events = runner.run(user_id=user_id, session_id=session.id, new_message=content)
#105
#106	# Extract final response
#107	for event in events:
#108	if event.is_final_response():
#109	response = event.content.parts[0].text
#110
#111	return response
#112
#113	return "No response generated"
#114
#115	# Example usage
#116	if __name__ == "__main__":
#117	response = asyncio.run(chat_with_agent(
#118	"I love Italian food and I'm planning a trip to Rome next month",
#119	user_id="alice"
#120	))
#121	print(response)
#122	```
#123
#124	## Multi-Agent Hierarchy with Shared Memory
#125
#126	Create specialized agents in a hierarchy that share memory:
#127
#128	```python
#129	from google.adk.tools.agent_tool import AgentTool
#130
#131	# Travel specialist agent
#132	travel_agent = Agent(
#133	name="travel_specialist",
#134	model="gemini-2.0-flash",
#135	instruction="""You are a travel planning specialist. Use search_memory to
#136	understand the user's travel preferences and history before making recommendations.
#137	After providing advice, use save_memory to save travel-related information.""",
#138	description="Specialist in travel planning and recommendations",
#139	tools=[search_memory, save_memory]
#140	)
#141
#142	# Health advisor agent
#143	health_agent = Agent(
#144	name="health_advisor",
#145	model="gemini-2.0-flash",
#146	instruction="""You are a health and wellness advisor. Use search_memory to
#147	understand the user's health goals and dietary preferences.
#148	After providing advice, use save_memory to save health-related information.""",
#149	description="Specialist in health and wellness advice",
#150	tools=[search_memory, save_memory]
#151	)
#152
#153	# Coordinator agent that delegates to specialists
#154	coordinator_agent = Agent(
#155	name="coordinator",
#156	model="gemini-2.0-flash",
#157	instruction="""You are a coordinator that delegates requests to specialist agents.
#158	For travel-related questions (trips, hotels, flights, destinations), delegate to the travel specialist.
#159	For health-related questions (fitness, diet, wellness, exercise), delegate to the health advisor.
#160	Use search_memory to understand the user before delegation.""",
#161	description="Coordinates requests between specialist agents",
#162	tools=[
#163	AgentTool(agent=travel_agent, skip_summarization=False),
#164	AgentTool(agent=health_agent, skip_summarization=False)
#165	]
#166	)
#167
#168	def chat_with_specialists(user_input: str, user_id: str) -> str:
#169	"""
#170	Handle user input with specialist agent delegation and memory.
#171
#172	Args:
#173	user_input: The user's message
#174	user_id: Unique identifier for the user
#175
#176	Returns:
#177	The specialist agent's response
#178	"""
#179	session_service = InMemorySessionService()
#180	session = session_service.create_session(
#181	app_name="specialist_system",
#182	user_id=user_id,
#183	session_id=f"session_{user_id}"
#184	)
#185	runner = Runner(agent=coordinator_agent, app_name="specialist_system", session_service=session_service)
#186
#187	content = types.Content(role='user', parts=[types.Part(text=user_input)])
#188	events = runner.run(user_id=user_id, session_id=session.id, new_message=content)
#189
#190	for event in events:
#191	if event.is_final_response():
#192	response = event.content.parts[0].text
#193
#194	# Store the conversation in shared memory
#195	conversation = [
#196	{"role": "user", "content": user_input},
#197	{"role": "assistant", "content": response}
#198	]
#199	mem0.add(conversation, user_id=user_id)
#200
#201	return response
#202
#203	return "No response generated"
#204
#205	# Example usage
#206	response = chat_with_specialists("Plan a healthy meal for my Italy trip", user_id="alice")
#207	print(response)
#208	```
#209
#210
#211
#212	## Quick Start Chat Interface
#213
#214	Simple interactive chat with memory and Google ADK:
#215
#216	```python
#217	def interactive_chat():
#218	"""Interactive chat interface with memory and ADK"""
#219	user_id = input("Enter your user ID: ") or "demo_user"
#220	print(f"Chat started for user: {user_id}")
#221	print("Type 'quit' to exit")
#222	print("=" * 50)
#223
#224	while True:
#225	user_input = input("\nYou: ")
#226
#227	if user_input.lower() == 'quit':
#228	print("Goodbye! Your conversation has been saved to memory.")
#229	break
#230	else:
#231	response = chat_with_specialists(user_input, user_id)
#232	print(f"Assistant: {response}")
#233
#234	if __name__ == "__main__":
#235	interactive_chat()
#236	```
#237
#238	## Key Features
#239
#240	### 1. Memory-Enhanced Function Tools
#241	- Function Tools: Standard Python functions that can search and save memories
#242	- Tool Context: Access to session state and memory through function parameters
#243	- Structured Returns: Dictionary-based returns with status indicators for better LLM understanding
#244
#245	### 2. Multi-Agent Memory Sharing
#246	- Agent-as-a-Tool: Specialists can be called as tools while maintaining shared memory
#247	- Hierarchical Delegation: Coordinator agents route to specialists based on context
#248	- Memory Categories: Store interactions with metadata for better organization
#249
#250	### 3. Flexible Memory Operations
#251	- Search Capabilities: Retrieve relevant memories through conversation history
#252	- User Segmentation: Organize memories by user ID
#253	- Memory Management: Built-in tools for saving and retrieving information
#254
#255	## Configuration Options
#256
#257	Customize memory behavior and agent setup:
#258
#259	```python
#260	# Configure memory search with filters
#261	# For Platform API, all filters including user_id go in filters object
#262	memories = mem0.search(
#263	query="travel preferences",
#264	filters={
#265	"AND": [
#266	{"user_id": "alice"},
#267	{"categories": {"contains": "travel"}}
#268	]
#269	},
#270	limit=5
#271	)
#272
#273	# Configure agent with custom model settings
#274	agent = Agent(
#275	name="custom_agent",
#276	model="gemini-2.0-flash", # or use LiteLLM for other models
#277	instruction="Custom agent behavior",
#278	tools=[memory_tools],
#279	# Additional ADK configurations
#280	)
#281
#282	# Use Google Cloud Vertex AI instead of AI Studio
#283	os.environ["GOOGLE_GENAI_USE_VERTEXAI"] = "True"
#284	os.environ["GOOGLE_CLOUD_PROJECT"] = "your-project-id"
#285	os.environ["GOOGLE_CLOUD_LOCATION"] = "us-central1"
#286	```
#287
#288	<CardGroup cols={2}>
#289	<Card title="Healthcare Agent Cookbook" icon="heart-pulse" href="/cookbooks/integrations/healthcare-google-adk">
#290	Build HIPAA-compliant healthcare agents with Google ADK
#291	</Card>
#292	<Card title="OpenAI Agents SDK" icon="cube" href="/integrations/openai-agents-sdk">
#293	Compare with OpenAI's agent framework
#294	</Card>
#295	</CardGroup>
#296
#297

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public